dataset.py

import os
import datasets

class ForestSegmentationDataset(datasets.GeneratorBasedBuilder):
    def _info(self):
        return datasets.DatasetInfo(
            features=datasets.Features({
                "sample_id": datasets.Value("string"),
                "image": datasets.Image(),  # Path to all_bands.tif
                "mask": datasets.Image(),   # Path to mask.tif
            }),
        )

    def _split_generators(self, dl_manager):
        # Streaming mode: use iter_files to lazily iterate over image files
        image_paths = dl_manager.iter_files("images")
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"image_paths": image_paths}
            )
        ]

    def _generate_examples(self, image_paths):
        for image_path in image_paths:
            if not image_path.endswith("all_bands.tif"):
                continue

            # Extract sample_id from path: images/0000005/<timestamp>/all_bands.tif
            parts = image_path.split("/")
            if len(parts) < 3:
                continue  # Skip malformed paths

            sample_id = parts[-3]  # "0000005"
            mask_path = f"masks/{sample_id}/mask.tif"

            yield sample_id, {
                "sample_id": sample_id,
                "image": image_path,
                "mask": mask_path,
            }